It’s no secret that artificial intelligence is actively used in writing music. But this process is completely incomprehensible to the average music lover, who does not understand what algorithms the program can use to compose music.
To explain how artificial intelligence manages to create music from scratch, let’s take a very simplified example. Let’s imagine that we want to develop an AI to generate songs in the style of Pink Floyd. The first thing to do is to “feed” as many of the group’s songs as possible to the artificial intelligence: this will allow the algorithm to identify repeating patterns.
Once these repeating patterns are developed, the program can figure out which note sequences and which rhythms are closest to the Pink Floyd style (based on data entered into its database) and which are less so. All of this information is then translated into a series of mathematical formulas that allow you to create a custom recipe to recreate the sound and mood of those specific performers. Thus, it is a method that is based on errors and is able to correct itself every time in order to become more and more efficient – the so-called machine learning.
Sounds like science fiction? Maybe, but if you think about it, we humans do the same: all musicians are inspired by other musicians, and the composition of a piece of music is the result of many harmonic, rhythmic, and melodic attempts and studies. The fact is that what we can learn or improve in years, artificial intelligence can do in a few hours. In addition, we must consider that usually, except in special cases, composer musicians specialize in a musical genre, and AI with the appropriate data can create jazz, rap, rock, metal, classical, house, reggae, and so on.
First of all, we have to keep in mind that at the moment AI is more of a tool than a real musician, as it is based on already existing data and therefore cannot create something truly innovative. For this reason, one of the main purposes for which it is used is as a source of inspiration: tens of millions of songs have been created to date, and it can not be easy to always have new ideas.
Help can come from technology: since artificial intelligence is able to create songs that reflect our needs, in fact, it is an unlimited source of ideas, which can then be taken, reworked and modified at will. A perfect example of this type is artist Holly Herendon, who duets with an artificial version of herself in the song Eternal.
Moving on to the world of film music, a possible application is soundtracks. In fact, it is possible to create symphonic works by starting with AI and then playing them with a real orchestra. The same applies to the advertising sector, which may use these songs to replace the standard, often impersonal and generic ones. Another example would be video games: usually a video game guarantees many hours of play, but conversely, its soundtrack lasts only a few hours. With AI, an even more immersive experience could be created with always new music able to adapt to the different phases of the game and the player’s style.